KGQA-1

This model is a fine-tuned version of google/flan-t5-large on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 2.0784
• Rouge1: 72.8963
• Rouge2: 60.8929
• Rougel: 69.6657
• Rougelsum: 72.9329
• Gen Len: 4.8819
• F1: 0.7593
• Recall: 0.7681
• Precision: 0.7508

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 0.001
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine_with_restarts
• num_epochs: 8

Training results

Training Loss	Epoch	Step	Validation Loss	Rouge1	Rouge2	Rougel	Rougelsum	Gen Len	F1	Recall	Precision
2.8587	1.0	598	2.2931	49.5203	26.8249	43.3252	49.5005	4.6943	0.5633	0.5546	0.5723
1.7685	2.0	1196	1.6857	52.6345	31.7615	46.5617	52.5831	4.7965	0.619	0.6295	0.6088
0.8979	3.0	1794	1.3095	65.3839	49.1969	60.9907	65.2835	4.8928	0.6898	0.6806	0.6992
0.4881	4.0	2392	1.4524	68.0576	53.7819	64.3964	67.9986	4.835	0.7239	0.7106	0.7378
1.2094	5.0	2990	3.2070	18.934	4.1916	14.7003	18.9198	6.0159	0.0005	0.001	0.0003
0.7018	6.0	3588	1.3772	68.1255	54.2242	64.3339	68.1513	4.7588	0.7125	0.69	0.7366
0.3275	7.0	4186	1.5585	72.2516	60.2665	68.9117	72.2482	4.9246	0.7643	0.7827	0.7468
0.112	8.0	4784	2.0784	72.8963	60.8929	69.6657	72.9329	4.8819	0.7593	0.7681	0.7508

Framework versions

• Transformers 4.43.3
• Pytorch 2.3.1+cu121
• Datasets 2.20.0
• Tokenizers 0.19.1